The noise suppression of 3D seismic data acquired from Junggar Basin based on deep learning
-
摘要: 【目的】 准噶尔盆地是我国重要的含油气盆地，其勘探目标已进入深层。该盆地的复杂近地表条件、勘探目标深度以及“两宽一高”的三维地震数据采集方式导致地震资料信噪比低、数据量大，这些问题对勘探目标的落实产生了影响。因此，压制噪声并提高三维地震资料的品质对于实现勘探目标至关重要。【方法】 随着深度学习理论的发展和硬件性能的提升，深度神经网络的学习能力和处理效率得到了显著提高。为此，基于残差学习和批归一化技术，构建了三维去噪卷积神经网络(Three-dimensional denoising convolutional neural network，3D-DnCNN)，并开发了适用于准噶尔盆地的基于深度学习的三维地震资料噪声压制流程。【结果和结论】 针对准噶尔盆地某大连片工区的实际需求，选取了覆盖次数高、信噪比高的区域的噪声压制结果构建高质量标签，并将训练好的3D-DnCNN网络应用于整个工区。研究结果表明，与常规工业流程相比，所提方法得到的同相轴一致性更好、断裂保持更完整、石炭系顶界与内幕更加清晰。此外，3D-DnCNN网络在高信噪比区域学习到的偏移画弧噪声特征，使其在整个工区的偏移画弧噪声压制能力优于常规工业流程。通过调整网络参数（如网络深度、卷积核大小及训练样本选择策略）可以进一步优化网络以适应不同地区的地震资料，从而增强了地震噪声压制技术的适用性和有效性。Abstract: [Objective] The Junggar Basin is an important oil- and gas-bearing basin in China, and the exploration targets have extended into deeper layers. The complex near-surface conditions, exploration target depth, and new 3D seismic data acquisition methods (the wide-azimuth, wide-bandwidth, and high-density data acquisition) in the basin have led to some challenges, such as low signal-to-noise ratio (SNR) and large data volumes, which affect the finding of exploration goals. Therefore, suppressing noise and improving the quality of 3D seismic data within the basin is crucial for finding the exploration goals. [Method] In recent years, the rapid development of deep learning theory has resulted in the great learning capabilities of deep neural networks, while the significant enhancement of hardware performance has enabled higher processing efficiency. Based on residual learning and batch normalization techniques, this study developed a 3D denoising convolutional neural network (3D-DnCNN) and proposed a deep learning-based 3D seismic data noise suppression workflow suited to the Junggar Basin. [Results and Conclusion] To suppress the noise for the dataset from a large area in the Junggar Basin, this paper selected noise suppression results from high-coverage, high-SNR areas to construct high-quality labels and applied the trained 3D-DnCNN network to the entire area. Compared with the results of conventional industrial workflows, the results of this study show better event consistency, intact fault preservation, and clearer top boundaries and internal features of the Carboniferous strata. Moreover, since the 3D-DnCNN network learned the characteristics of migration arch noise in high-SNR areas, its ability to suppress such noise across the entire area was also significantly superior to conventional industrial workflows. By adjusting network parameters (such as network depth, convolution kernel size, and training sample selection strategy), the network can be further optimized to adapt to seismic data from different regions, thereby enhancing the applicability and effectiveness of seismic noise suppression techniques.
-
Keywords:
- Junggar Basin /
- deep learning /
- convolutional neural network /
- noise suppression
-
-
[1] 陈文超，刘达伟，魏新建，等. 基于地震资料有效信息约束的深度网络无监督噪声压制方法[J]. 煤田地质与勘探，2021，49(1)：249-256. CHEN Wenchao，LIU Dawei，WEI Xinjian，et al. Unsupervised noise suppression method for depth network seismic data based on prior information constraint[J]. Coal Geology & Exploration，2021，49(1)：249-256.
[2] 刘法启，张关泉. 小波变换与F-K算法在滤波中的应用[J]. 石油地球物理勘探，1996，31(6)：782-791. LIU Faqi，ZHANG Guanquan. Application of wavelet transform and F-K algorithm in filtering[J]. Oil Geophysical Prospecting，1996，31(6)：782-791.
[3] 康冶，于承业，贾卧，等. f-x域去噪方法研究[J]. 石油地球物理勘探，2003，38(2)：136-138. KANG Ye，YU Chengye，JIA Wo，et al. A study on noise-suppression method in f-x domain[J]. Oil Geophysical Prospecting，2003，38(2)：136-138.
[4] MALLAT S G. A theory for multiresolution signal decomposition：The wavelet representation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence，1989，11(7)：674-693.
[5] TANG Na，ZHAO Xian，LI Yue，et al. Adaptive threshold shearlet transform for surface microseismic data denoising[J]. Journal of Applied Geophysics，2018，153：64-74.
[6] LIU Cai，CHEN Changle，WANG Dian，et al. Seismic dip estimation based on the two-dimensional Hilbert transform and its application in random noise attenuation[J]. Applied Geophysics，2015，12(1)：55-63.
[7] 金雷，李月，杨宝俊. 用时频峰值滤波方法消减地震勘探资料中随机噪声的初步研究[J]. 地球物理学进展，2005，20(3)：724-728. JIN Lei，LI Yue，YANG Baojun. Reduction of random noise for seismic data by time-frequency peak filtering[J]. Progress in Geophysics，2005，20(3)：724-728.
[8] DUNCAN G，BERESFORD G. Median filter behaviour with seismic data 1[J]. Geophysical Prospecting，1995，43(3)：329-345.
[9] CHEN Yangkang，ZU Shaohuan，WANG Yufeng，et al. Deblending of simultaneous source data using a structure-oriented space-varying Median filter[J]. Geophysical Journal International，2020，222(3)：1805-1823.
[10] PORSANI M J，URSIN B，SILVA M G，et al. Dip-adaptive singular-value decomposition filtering for seismic reflection enhancement[J]. Geophysical Prospecting，2013，61(1)：42-52.
[11] 马继涛，王建花，刘国昌. 基于频率域奇异值分解的地震数据插值去噪方法研究[J]. 石油物探，2016，55(2)：205-213. MA Jitao，WANG Jianhua，LIU Guochang. Seismic data noise attenuation and interpolation using singular value decomposition in frequency domain[J]. Geophysical Prospecting for Petroleum，2016，55(2)：205-213.
[12] HE Kaiming，ZHANG Xiangyu，REN Shaoqing，et al. Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas，NV，USA. IEEE，2016：770-778.
[13] RUSSAKOVSKY O，DENG Jia，SU Hao，et al. ImageNet large scale visual recognition challenge[J]. International Journal of Computer Vision，2015，115(3)：211-252.
[14] BELINKOV Y，GEHRMANN S，PAVLICK E. Interpretability and analysis in neural NLP[C]//Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics：Tutorial Abstracts. Online. Stroudsburg，PA，USA：Association for Computational Linguistics，2020：5.
[15] RADFORD A, NARASIMHAN K, SALIMANS T, et al. Improving Language Understanding by Generative Pre-Training[J]. OpenAI blog, 2018.
[16] BROWN T, MANN B, RYDER N, et al. Language Models are Few-Shot Learners[J]. Advances in Neural Information Processing Systems, 2020, 33: 1877-1901.
[17] CHEN Chenyi，SEFF A，KORNHAUSER A，et al. DeepDriving：Learning affordance for direct perception in autonomous driving[C]//2015 IEEE International Conference on Computer Vision (ICCV). Santiago，Chile. IEEE，2015：2722-2730.
[18] XUE Qingwen，WANG Ke，LU J J，et al. Rapid driving style recognition in car-following using machine learning and vehicle trajectory data[J]. Journal of Advanced Transportation，2019，2019：9085238.
[19] CHEN Jie，WU Zhongcheng，ZHANG Jun. Driving safety risk prediction using cost-sensitive with nonnegativity-constrained autoencoders based on imbalanced naturalistic driving data[J]. IEEE Transactions on Intelligent Transportation Systems，2019，20(12)：4450-4465.
[20] PEREIRA S，PINTO A，ALVES V，et al. Brain tumor segmentation using convolutional neural networks in MRI images[J]. IEEE Transactions on Medical Imaging，2016，35(5)：1240-1251.
[21] 于子叶，储日升，盛敏汉. 深度神经网络拾取地震P和S波到时[J]. 地球物理学报，2018，61(12)：4873-4886. YU Ziye，CHU Risheng，SHENG Minhan. Pick onset time of P and S phase by deep neural network[J]. Chinese Journal of Geophysics，2018，61(12)：4873-4886.
[22] ROSS Z E，YUE Yisong，MEIER M A，et al. PhaseLink：A deep learning approach to seismic phase association[J]. Journal of Geophysical Research：Solid Earth，2019，124(1)：856-869.
[23] 许祥，邹志辉，韩明亮，等. 联合地震初至走时与早至波形的深度学习速度建模[J]. 地球物理学报，2023，66(12)：5107-5122. XU Xiang，ZOU Zhihui，HAN Mingliang，et al. Deep-learning velocity model building by jointly using seismic first arrivals and early-arrival waveforms[J]. Chinese Journal of Geophysics，2023，66(12)：5107-5122.
[24] DUAN Xudong，ZHANG Jie. Reflection residual statics estimated using a high-resolution neural network[J]. Geophysics，2022，87(4)：U121-U134.
[25] CHOI Y，JO Y，SEOL S J，et al. Deep learning spectral enhancement considering features of seismic field data[J]. Geophysics，2021，86(5)：V389-V408.
[26] CHEVITARESE D，SZWARCMAN D，SILVA R M D，et al. Seismic facies segmentation using deep learning[C]//2018 AAPG Annual Convention & Exhibition. Salt Lake City，Utah，USA. Tulsa，OK，USA：American Association of Petroleum Geologists，2018：1-16.
[27] ZHANG Guoyin，WANG Zhizhang，CHEN Yangkang. Deep learning for seismic lithology prediction[J]. Geophysical Journal International，2018，215(2)：1368-1387.
[28] YUAN Sanyi，LIU Jiwei，WANG Shangxu，et al. Seismic waveform classification and first-break picking using convolution neural networks[J]. IEEE Geoscience and Remote Sensing Letters，2018，15(2)：272-276.
[29] WU Xinming，LIANG Luming，SHI Yunzhi，et al. FaultSeg3D：Using synthetic data sets to train an end-to-end convolutional neural network for 3D seismic fault segmentation[J]. Geophysics，2019，84(3)：IM35-IM45.
[30] ZHANG Yinxue，TIAN Xuemin，DENG Xiaogang，et al. Seismic denoising based on modified BP neural network[C]//2010 Sixth International Conference on Natural Computation. Yantai，China. IEEE，2010：1825-1829.
[31] KLOCHIKHINA E，CRAWLEY S，FROLOV S，et al. Leveraging deep learning for seismic image denoising[J]. First Break，2020，38(7)：41-48.
[32] ZHANG Kai，ZUO Wangmeng，CHEN Yunjin，et al. Beyond a Gaussian denoiser：Residual learning of deep CNN for image denoising[J]. IEEE Transactions on Image Processing，2017，26(7)：3142-3155.
[33] TRICKETT S. F-xy Cadzow noise suppression[C]//SEG Technical Program Expanded Abstracts 2008. Society of Exploration Geophysicists，2008：2586-2590.
[34] FEHMERS G C，HÖCKER C F W. Fast structural interpretation with structure-oriented filtering[J]. Geophysics，2003，68(4)：1286.
计量
- 文章访问数: 28
- HTML全文浏览量: 1
- PDF下载量: 8
下载:
